JP2012226705A - Structure processing method of non-structured data - Google Patents

Abstract

PROBLEM TO BE SOLVED: To convert non-structured data such as text into a form of structured data by extracting and polishing useful information from the non-structured data.SOLUTION: A structure processing method of non-structured data uses a data structuring device for converting non-structured data contained in structured data relating to navigation services into structured data. A processing procedure using the data structuring device includes the step of extracting item information that is non-structured data relating to a predetermined item from the structured data relating to the navigation services, an information element extraction step of extracting information elements contained in the item information in accordance with the description pattern of information elements and the description order of information elements contained in the item information, and the step of converting the extracted information elements into structured data.

Description

  The present invention relates to a structured processing method for unstructured data, and more particularly to a structured processing method for unstructured data that extracts useful information from unstructured data such as text, refines it, and converts it into structured data. .

  In recent years, with the dramatic improvement in performance and multi-functionality of mobile terminals, represented by the emergence of smartphones, stores (such as restaurants and convenience stores) and facilities (such as transportation facilities and hotels) around the current location and destination An information service for guiding information (hereinafter referred to as a navigation service) can be used regardless of time and place. For example, the user can receive information regarding the store or facility on site immediately before using the store or facility.

  Further, an API (Application Program Interface) that publishes a database (hereinafter referred to as store information DB) that stores store and facility information that is the core of a system that provides a navigation service to the outside is also provided (for example, non-patent literature). 1 and 2). In these APIs, for example, as shown in FIG. 14, various information related to the navigation service may be obtained, and a third-party provider can newly devise and provide a unique navigation service while accessing the existing store information DB. There is room.

“Hot Pepper Web Service VerUp, 570,000 Store Name APIs Released & Latitude / Longitude / Details Added!”, Tatami Lab, [online], [Search February 10, 2011], Internet <http: //www.tatamilab. jp / rnd / archives / 000375.html> “GourNavi releases API, what is behind it?”, @IT, [online], [February 10, 2011 search], Internet <http://jibun.atmarkit.co.jp/lskill01/special /gr_api/gr_api01.html>

  Here, when using the navigation service from a mobile terminal, the display capability (screen size) of the terminal is limited, and in order to save the user's effort as much as possible, only information in line with the user's intention and purpose It is desirable that the navigation result or the like is displayed in a form that is selected / emphasized.

  In the conventional store information DB, store information is structured data such as XML, and each item information of the store information (for example, an access method, an address, a business schedule, etc.) is defined by a predetermined XML tag. However, the item information itself is stored in a text format, that is, as unstructured data. For example, regarding the access method, information such as the nearest station name and station exit is not structured.

  In a navigation service using a portable terminal, various modes can be considered for selection / emphasis according to the user's intention and purpose. In order to realize such selection / emphasis, it is useful for selection / emphasis. It is desirable to prepare available information as structural type data in advance.

  In view of such a problem, an object of the present invention is to extract the structure of unstructured data from the unstructured data included in the structured data related to the navigation service, refine it, and convert it into structured data. It is to provide a processing method.

  The structured processing method of unstructured data according to the present invention is a structured processing method of unstructured data by a data structuring apparatus that converts unstructured data included in structured data related to a navigation service into structured data. The process procedure by the data structuring apparatus extracts item information that is unstructured data related to a predetermined item from structured data related to the navigation service, and description of information elements included in the item information According to the description order of the pattern and the information element, an information element extraction step for extracting the information element included in the item information and a step for converting the extracted information element into structural type data are included.

  According to the structured processing method for unstructured data according to the present invention, useful information can be extracted from the unstructured data included in the structured data related to the navigation service, refined, and converted into structured data. It becomes.

FIG. 1 is a diagram showing a schematic configuration of a system including a data structuring apparatus according to an embodiment of the present invention. FIG. 2 is a diagram showing a schematic configuration of a data structuring apparatus according to an embodiment of the present invention. FIG. 3 is a diagram illustrating an example of items including useful unstructured data. FIG. 4 is a diagram illustrating an example of data structuring of item information related to the access method. FIG. 5 is a flowchart for acquiring an information element related to a station name from item information related to an access method. FIG. 6 is a flowchart for acquiring an information element related to an exit from item information related to an access method. FIG. 7 is a flowchart for acquiring information elements relating to movement information from item information relating to access methods. FIG. 8 is a diagram illustrating an example of item information including a plurality of access methods. FIG. 9 is a diagram illustrating an example of data structuring of item information related to an address. FIG. 10 is a flowchart for acquiring information elements from item information relating to addresses. FIG. 11 is a diagram illustrating an example of the data structure of item information related to a business schedule. FIG. 12 is a flowchart for acquiring information elements from item information related to a business schedule. FIG. 13 is a diagram illustrating an example of a synonym list related to a business schedule. FIG. 14 is a diagram showing an example of structured data relating to a conventional navigation service.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 is a diagram showing a configuration of an entire system including a data structuring apparatus according to an embodiment of the present invention. The data structuring apparatus 1 includes a data acquisition unit 10, a data structuring unit 20, and a store metadata DB 30. The external server device 40 connected to the data structuring device 1 includes a morphological analysis unit 41 and a station name DB 42. In the present embodiment, the morphological analysis unit 41 and the station name DB 42 are described as functions of the external server device 40. However, the data structuring apparatus 1 has the functions of the morpheme analysis unit 41 and the station name DB 42. Note that it is also good.

  The data acquisition unit 10 acquires store information described by item from an external store information DB using an API provided by an existing navigation service. Here, the store information is structured type data such as an XML document, for example, and the item of store information indicates, for example, a predetermined XML tag (<address> <access> etc.) included in the XML document. is there. Hereinafter, data relating to each item (for example, data defined by an XML tag) is referred to as item information.

  The data structuring unit 20 extracts items including useful unstructured data from the store information acquired by the data acquisition unit 10, and assigns useful information elements (for example, meta information) from the item information regarding the items. Get possible numbers and words). Here, although the details will be described later, the data structuring unit 20 is included in the item information in accordance with a character string search by a regular expression corresponding to the description pattern of the information element, the description order of the information elements in the item information, or the like. Extract information elements. In addition, the data structuring unit 20 uses the morphological analysis unit 41 and the station name DB 42 of the external server device 40 to use characteristic words (for example, “ Information elements included in the item information can be extracted by searching for a character string consisting of words such as “station” and “mouth” that are useful for information element search. The data structuring unit 20 converts the extracted information element into structural type data (metadata such as RDF: Resource Description Framework) suitable for mechanical data interpretation such as Semantic Web, and stores it in the store metadata DB 30. . Here, converting an information element into metadata indicates that predetermined meta information is given to each information element, for example, each information element is defined by a predetermined XML tag.

  The store metadata DB 30 stores the metadata and provides an API related to access to the metadata to an external third party. An external third-party provider acquires necessary metadata using an API provided by the store metadata DB 30 and selects / emphasizes the client terminal as a Web service, for example, according to the user's intention and purpose. A navigation service can be provided.

  FIG. 2 is a diagram showing the configuration of the data structuring apparatus 1 shown in FIG. 1 in more detail. As described above, the data structuring apparatus 1 includes the data acquisition unit 10, the data structuring unit 20, and the store metadata DB 30, and the data structuring unit 20 includes the data extraction unit 21, the data refinement unit 22, The data converter 23 is provided. The processes of the data acquisition unit 10 and the store metadata DB 30 are as described above, and hereinafter, each process of the data extraction unit 21, the data refinement unit 22, and the data conversion unit 23 included in the data structuring unit 20 will be described in detail. .

  The data extraction unit 21 extracts items including useful unstructured data from the store information acquired by the data acquisition unit 10. For example, items containing useful unstructured data can be defined offline in advance. FIG. 3 is a diagram illustrating an example of items including useful unstructured data. FIG. 3A shows items related to “access method”, and item information is defined by an <access> tag. As shown in the figure, item information defined by the <access> tag is unstructured data, but includes useful information elements such as a station name, an exit, a movement method, a movement time, and a time unit. FIG. 3B shows items related to “address”, and item information is defined by an <address> tag. As shown in the drawing, the item information defined by the <address> tag is unstructured data, but includes useful information elements such as prefectures, municipalities, street addresses, building names, and floor numbers. FIG. 3C shows items related to “business schedule”, and item information is defined by an <open> tag. As illustrated, the item information defined by the <open> tag is unstructured data, but includes useful information elements such as business days and business hours.

  Even when the data structure (item name, for example, the name of the XML tag) varies between different store information DBs, the data extraction unit 21 absorbs such variations by referring to, for example, the XML schema definition. Can do. When the data extraction unit 21 extracts items including useful unstructured data, the data extraction unit 21 supplies item information included in the items to the data refinement unit 22.

  The data refiner 22 extracts useful information elements from the item information included in the items extracted by the data extractor 22. The information element is, for example, a numerical value or a word having a certain meaning, and in the case of the access method of FIG. 3A, the station name, exit, travel method, travel time, time unit, etc. of the address of FIG. In this case, in the case of the business schedule shown in FIG. 3C, such as a prefecture city, a street address, a building name, and a floor number, business days, business hours, and the like are listed as examples of information elements.

  The data refinement unit 22 extracts information elements included in the item information in accordance with a character string search using a regular expression corresponding to the description pattern of the information elements, a description order of the information elements in the item information, and the like. In addition, the data structuring unit 20 uses the morphological analysis unit 41 and the station name DB 42 of the external server device 40 to use characteristic words (for example, “ Information elements included in the item information can be extracted by searching for a character string including “station”, “mouth”, and the like. Details of the information element acquisition will be described later. The data refiner 22 supplies the acquired information element to the data converter 23.

  The data conversion unit 23 converts the information element supplied from the data refinement unit 22 into structural data (metadata) suitable for mechanical data interpretation such as Semantic Web. Specifically, the data conversion unit 23 converts each information element into structured data by defining each information element with a tag indicating the meaning (meta information) of each information element. Here, in the case of the access method of FIG. 3A, the tag indicating the meaning of each information element is <station> indicating the station name, <exit> indicating the exit, <transport> indicating the moving method, and the moving time. In the case of the address in Fig. 3 (b), such as <time> indicating <time> indicating the unit of time, <area> indicating the prefecture city, <number> indicating the address, <building> indicating the building name, and the floor number In the case of the business schedule shown in FIG. 3C, such as <floor>, <day> indicating business days, <hour> indicating business hours, and the like. The data conversion unit 23 stores the converted structural type data (metadata) in the store metadata DB 30.

  Hereinafter, acquisition of information elements from item information by the data refinement unit 22 will be described in detail. Specifically, the processing of the data refinement unit 22 will be described in detail in the order of the access method, address, and business schedule items shown in FIGS.

(A) Access method For obtaining information elements from item information related to access methods, the description content of item information related to access methods is categorized according to certain rules (such as the description pattern of each information element and the description order of each information element). Focus on what is being done. For example, the item information regarding the access method obtained from the API of Non-Patent Document 1 is described in the order of station name, exit, travel method, travel time, time unit, and the station name, exit, travel method, travel time, Information elements for each time unit are expressed by a common description pattern. Note that item information that is not categorized may exist for some access methods. However, for item information of such partial access methods, exception processing may be defined individually, which affects the effectiveness of the present invention. Note that it does not give

  FIG. 4 is a diagram illustrating an example of data structuring of item information related to the access method. For example, the item information “1-minute walk from Exit A / A Exit / 3-minute walk from Exit B” defined by the <access> tag is the information element related to the station name “A Station” and “Exit A” and “Exit B”. It contains multiple information elements: an information element related to the exit, an information element related to the transportation means “walk”, an information element related to the travel time “1” and “3”, and an information element related to the time unit “minute”. As shown in the graph of FIG. Hereinafter, acquisition of each information element of the station name, exit, movement method, movement time, and time unit by the data refinement unit 22 will be described in detail.

  FIG. 5 is a flowchart for acquiring an information element related to a station name from item information related to an access method. The data refinement unit 22 removes all symbols for emphasizing the station name such as “”, [], <> as preprocessing (step S 101), and uses the morpheme of the external server device 40 for the item information that has been preprocessed. This is provided to the analysis unit 41. Next, the data refinement unit 22 performs station name extraction based on the analysis result received from the morpheme analysis unit 41 (step S102). Specifically, the data refinement unit 22 searches for a character string including a proper noun obtained by morphological analysis of item information and “station” which is a characteristic word based on the description pattern of the information element. The data refinement unit 22 searches for a word “proprietary noun +“ station ”” and, when a corresponding word is detected, extracts the word as an information element related to the station name.

  Further, for example, when the information element relating to the station name is not detected by the morphological analysis (eg, “Mitaka station” is abbreviated as “Mitaka”) (No in step S103), the data refinement unit 22 determines the station name of the external server device 40. The DB 42 is collated to extract information elements related to the station name (step S104). Specifically, the data refinement unit 22 compares the proper noun obtained from the item information by the morphological analysis and the station name stored in the station name DB 42. For example, if the data refinement unit 22 acquires information such as the prefecture of the store, the data refinement unit 22 can narrow the search range of the station name DB 42 using the information. Further, when a plurality of station names are detected as candidates from the station name DB 42, the data refinement unit 22 can select one station name as an information element based on the longest match of the character string of the proper noun and the station name. . Note that the comparison between proper nouns and station names and the selection of station names by the longest match method are not performed by the data refinement unit 22 but by the station name DB 42, and the data refinement unit 22 receives only the selection result from the station name DB 42. Is possible.

FIG. 6 is a flowchart for acquiring an information element related to an exit from item information related to an access method. The data refinement unit 22 removes all symbols for emphasizing the exit, such as “”, [], <>, as preprocessing (step S 201). Next, the data refinement unit 22 extracts an information element related to the exit included in the item information using a regular expression corresponding to the description pattern of the information element related to the exit (step S202). For example, the data refinement unit 22 can extract information elements related to the exit using the following regular expression. It should be noted that the following regular expressions are merely examples, and those skilled in the art can create various regular expressions separately from the following regular expressions.
"[AZ] \\ d * (exit | exit) | (east | west | south | north | center) exit | \\ d (exit | exit)"

  Further, for example, when an information element related to the exit is not detected by a regular expression such as an exit represented by “proper noun +“ mouth ”” such as “Yaesuguchi” (No in step S203), the data refinement unit 22 Is searched for a character string consisting of a proper noun obtained by morphological analysis and “mouth” which is a characteristic word based on the description pattern of the information element (step S204). The data refinement unit 22 searches for a word “proprietary noun +“ mouth ”” and, when a corresponding word is detected, extracts the word as an information element related to the exit (step S204). As the result of the morpheme analysis used here, for example, the analysis result acquired in step S102 regarding the extraction of the station name may be used, or the item information is transmitted again to the morpheme analysis unit 41 to newly acquire the result of the morpheme analysis. You may do it.

FIG. 7 is a flowchart for acquiring information elements relating to movement information (movement method, movement time, and time unit) from item information relating to the access method. The data refinement unit 22 removes all symbols for emphasizing movement information such as “”, [], <> as preprocessing (step S 301). Next, the data refiner 22 extracts the movement information from the item information using a regular expression corresponding to the movement information description pattern including a plurality of information elements (step S302). For example, the data refinement unit 22 can extract movement information including a plurality of information elements by the following regular expression. It should be noted that the following regular expressions are merely examples, and those skilled in the art can create various regular expressions separately from the following regular expressions.
"((Walk | train | bus | \\ b * (minute | second)) | \\ b * minute"

  When acquiring the movement information including a plurality of information elements, the data refinement unit 22 acquires information elements relating to the movement method, the movement time, and the time unit based on the order of description in the movement information (step S303). For example, since the travel time is generally written in Arabic numerals, the data refinement unit 22 extracts the travel time based on the travel time (first information element) description pattern, and then in the travel information (item information). Depending on the description order of the information elements, it is possible to extract the movement method and time unit as other information elements from the movement time as a base point.

  Although the method of acquiring each information element from the item information related to the access method has been described with reference to FIGS. 5 to 7, a plurality of access methods may be described in one item information. FIG. 8 is a diagram illustrating an example of item information including a plurality of access methods. Even in the case of FIG. 8, the description of each access method is categorized according to a certain rule (the description pattern of each information element, the description order of each information element, etc.) as described above. That is, for example, item information regarding each access method is described in the order of station name, exit, travel method, travel time, time unit, and information elements for each station name, exit, travel method, travel time, time unit Are expressed in a common pattern. For this reason, the data refinement part 22 can acquire each information element of a some access method by repeating the process shown, for example in FIGS. In addition, for the second and subsequent access methods, even when there is an omitted information element among the station name, exit, movement method, movement time, and time unit, for example, the omitted first information element (for example, station name) The structural type data can be generated by associating the data described after (for example, exit and movement information) with the first information element (= station name) described immediately before. Specifically, for example, in the item information “A station A exit 1 minute walk / B exit 3 minutes walk” in FIG. 4, the station name information element (No. 1 information element) is omitted. In this case, the structure type data is created by associating “3 minutes on foot from Exit B”, which is the data described after the station name, with “A station”, which is the first information element (station name) described immediately before. can do.

(B) Address For acquiring information elements from item information related to addresses, the description contents of item information related to addresses are categorized according to certain rules (such as the description pattern of each information element and the description order of each information element). Focus on that. For example, general addresses are described in the order of prefectures, municipalities, addresses, building names, and floors, and information elements such as addresses and floors are expressed in a common description pattern. Note that item information that is not categorized may exist for some addresses, but for such item information, exception processing may be defined individually, which affects the effectiveness of the present invention. Note that it is not a thing.

  FIG. 9 is a diagram illustrating an example of data structuring of item information related to an address. For example, the item information “3-16-1 Sagamiono, Sagamihara City, Kanagawa Prefecture” defined by the <address> tag includes information elements related to prefectures and municipalities “Sagamiono, Sagamihara City, Kanagawa Prefecture” FIG. 9 includes a plurality of information elements including an information element related to the address “3-16-1”, an information element related to the building name “Regaro Building”, and an information element related to the floor number “B1” and “1F”. It can be structured as shown in the graph. Hereinafter, acquisition of each information element of the prefectural municipalities, street addresses, building names, and floors by the data refinement unit 22 will be described in detail.

  FIG. 10 is a flowchart for acquiring each information element from the item information regarding the address. Using the fact that the address is described using Arabic numerals, the data refinement unit 22 extracts the character string before the address (Arabic numerals) as an information element related to the prefectural municipalities (step S401). Note that information elements related to prefectures and municipalities can be subdivided into smaller information elements such as prefectural units and municipalities, for example, by existing methods. However, in the present invention, such subdivision is not described in detail. The data refinement unit 22 acquires a character string composed of Arabic numerals and symbols (for example, a hyphen “-”) following the information element related to the prefecture municipality as the information element related to the address (step S402). For example, when a plurality of Arabic numerals are written by hyphens, etc., it is possible to correspond to smaller information elements such as chome, number, and number in order from the first Arabic numeral.

  The data refinement unit 22 extracts a character string following the information element related to the address as the information element related to the building name (step S403). Since the information element related to the building name extracted in step S403 may include an information element related to the floor number of the building, the data refinement unit 22 uses a regular expression to express a character string indicating the floor number such as “floor” and “B”. (Step S404). It should be noted that when the information element related to the floor is extracted, the data refinement unit 22 excludes the character string corresponding to the information element related to the floor from the information element related to the building name.

(C) Business schedule When acquiring information elements from item information related to business schedules, the description content of item information related to business schedules is categorized according to certain rules (such as the description pattern of each information element and the order in which each information element is described). Focus on what is being done. For example, item information relating to a business schedule includes business days, business categories, opening times, closing times, LO (last order) times, and information on business days, business categories, opening times, closing times, LO times. Elements are expressed in a common description pattern. Hereinafter, for convenience of explanation, the expression “business hours” will be used as appropriate as a general term including opening hours, closing times, and LO times.

In particular, information element patterns (hereinafter referred to as information element patterns) included in item information relating to business schedules are roughly divided into the following three information element patterns. Information on information elements included in such item information is useful for extracting information elements.
(1) Information element pattern 1 (business hours only)
Example: 11:30 to 5:00
(2) Information element pattern 2 (first business day, followed by business hours)
Example: Monday to Saturday / 17: 00 to 23:30 (L.O. 23:00)
Sun / Holiday / 17: 30-23: 00 (L.O. 22:30)
(3) Information element pattern 3 (first business division, followed by business day, business hours)
Example: Lunch Mon-Fri 11: 30-14: 00
Dinner Mon-Sat 17: 00-23: 30, Sundays and holidays 16: 30-22: 30

  In addition, there may be item information that is not categorized for some business schedules. However, for such item business schedule information, it is sufficient to define exception handling individually, which affects the effectiveness of the present invention. Note that it does not give

  FIG. 11 is a diagram illustrating an example of the data structure of item information related to a business schedule. For example, “Lunch Monday-Friday 11: 00-15: 00 (LO14: 00) Dinner Monday-Friday 18: 00-23: 00 (LO22: 00) Saturdays, Sundays, and holidays 18:00 Item information of ˜22: 00 (LO21: 00) includes information elements related to business days “Monday” to “Sunday” and “holidays”, information elements related to business categories “lunch” and “dinner”, and “11 : 10:00, “18:00”, “19:00”, information elements related to the opening hours, “15:00”, “23:00”, “22:00”, information elements related to the closing times, “14:00”, “ It includes a plurality of information elements called information elements relating to the LO time of 22:00 and 21:00, and can be structured as shown in the graph of FIG. In FIG. 11, for convenience of explanation, the subsequent data structure is described only for “Tuesday” and “Sunday”, but “Monday”, “Wednesday”, “Thursday”, and “Friday” are similar to “Tuesday”. It should be noted that the data structure “Saturday” and “Holiday” have the same data structure as “Sunday”. Hereinafter, acquisition of each information element of the business day, business division, opening time, closing time, and LO time by the data refinement unit 22 will be described in detail.

  The data refinement unit 22 identifies the information element pattern of the item information related to the business schedule (step S501). As described above, information element patterns of item information related to business schedules are roughly divided into three patterns. The data refinement unit 22 can identify which information element pattern the item information corresponds to by checking the leading character string of the item information. Specifically, when the first character string of the item information indicates business hours (for example, Arabic numerals), the data refinement unit 22 identifies that the item information is the information element pattern 1, and the first character string of the item information When indicating business days (for example, Monday to Sunday, holidays), the item information is identified as information element pattern 2, and when the first character string of the item information indicates a business category (for example, lunch, dinner, etc.), item information Can be identified as the information element pattern 3.

  Next, the data refinement unit 22 extracts information elements included in the item information according to the description pattern of the information elements in the identified information element pattern and the description order of the information elements. First, the data refinement unit 22 extracts business hours from the item information by a character string search using a regular expression corresponding to the business hours description pattern (step S502). This is because business hours are included in all of the information element patterns 1 to 3. Here, the description pattern of sales information includes two description patterns of a format as a time zone “09: 0 to 22:00” and a format as a time “21:00”. Here, when the data refinement unit 22 detects a character string including a description pattern indicating a time zone such as “09: 0 to 22:00”, the data refinement unit 22 extracts the first time as an information element related to the opening time, and sets the last time. Extracted as information elements related to closing hours. Further, when the data refinement unit 22 detects a character string including a description pattern indicating a time such as “21:00”, the data refinement unit 22 extracts the time as an information element related to the LO time. The data refinement unit 22 extracts the LO time by combining a character string such as “last order”, “LO”, and “LO” with the time and detecting the character string corresponding to the combination. Information elements can be extracted.

  After the business hours are extracted, the data refinement unit 22 extracts each information element with a regular expression corresponding to the description pattern of the information element included in the identified information element pattern (step S503). Specifically, when the identified information element pattern is the information element pattern 2, the data refinement unit 22 extracts the information element related to the business day by a regular expression corresponding to the business day description pattern included in the information element pattern. To do. Further, when the identified information element pattern is the information element pattern 3, the data refinement unit 22 uses the regular expressions corresponding to the description patterns of the business segment and business day included in the information element pattern, respectively. Extract information elements about. Note that when the identified information element pattern is the information element pattern 1, the data refinement unit 22 ends the extraction of the information element because the information element pattern includes no information elements other than business hours.

  Regarding the extraction of business days and business divisions, the data refinement unit 22 can extract business days and business divisions using regular expressions that take into account synonyms for business schedules. FIG. 13 is a diagram illustrating an example of a synonym list related to a business schedule. By performing a character string search using a regular expression based on such synonyms, it becomes possible to more reliably extract information elements related to business days and business categories.

  In addition, regarding the extraction of the information element from the item information regarding the business schedule shown in steps S501 to S503, specific regular expressions are not illustrated, but those skilled in the art will describe the business category, business day, and business time. Note that various regular expressions can be created depending on the pattern.

  As described above, according to the present embodiment, the data structuring apparatus 1 determines the information elements included in the item information according to the description pattern of the information elements included in the item information that is unstructured data and the description order of the information elements. Is extracted, and the extracted information element is converted into structured data. This makes it possible to extract useful information from the unstructured data included in the structured data related to the navigation service, refine it, and convert it into structured data. Further, the data structuring apparatus 1 stores the structured data converted from the unstructured data in the store metadata DB 30. The third-party provider implements a new navigation service with the store metadata DB 30 as a core, thereby searching for store information according to a new usage form, that is, a new selection / emphasis according to the user's intention and purpose. It becomes possible to carry out with high precision. Also, the store metadata DB 30 is obtained by converting the format of the metadata extracted and refined for each information element from item information of unstructured data including useful information elements such as an access method, an address, and a business schedule into structural data. By accumulating data, it becomes possible to utilize the existing foundation of mechanical data interpretation, and to realize a new navigation service quickly and economically.

  Further, according to the present embodiment, the data structuring apparatus 1 can extract information elements included in the item information by performing a character string search using a regular expression corresponding to the description pattern of the information elements. This makes it possible to extract various information elements according to the description pattern using a common regular expression. Also, regular expressions can create flexible search conditions. For example, even if an information element has some fluctuations from the description pattern, it can be extracted with a common regular expression according to the design of the regular expression. Is possible.

  Further, according to the present embodiment, the data structuring apparatus 1 searches for a character string composed of a proper noun obtained by morphological analysis of item information and a feature word based on the description pattern of the information element, thereby obtaining the item Information elements included in the information can be extracted. This makes it possible to extract information elements that are not extracted by regular expressions. In addition, for example, in cooperation with an external database, information elements can be extracted by comparing the values of proper nouns obtained by morphological analysis and the external database.

  Further, according to the present embodiment, the data structuring apparatus 1 extracts the first information element based on the description pattern of a certain first information element, and then, according to the description order of the information elements in the item information, The extraction of other information elements can be performed using one information element as a base point. Thereby, for example, even if an information element is difficult to extract by regular expression or morphological analysis, it is possible to extract another information element as a base point.

  Further, according to the present embodiment, the data structuring apparatus 1 identifies the information element pattern of the item information, and includes it in the item information according to the description pattern of the information elements in the identified information element pattern and the description order of the information elements. Information elements can be extracted. Accordingly, each information element can be extracted after accurately grasping the information element included in the item information in advance, so that the information element can be extracted with high accuracy.

  Although the present invention has been described based on the drawings and examples, it should be noted that those skilled in the art can easily make various modifications and corrections based on the present disclosure. Therefore, it should be noted that these variations and modifications are included in the scope of the present invention. For example, the functions included in each component, each step, etc. can be rearranged so that there is no logical contradiction, and multiple components, steps, etc. can be combined or divided into one It is.

DESCRIPTION OF SYMBOLS 1 Data structuring apparatus 10 Data acquisition part 20 Data structuring part 21 Data extraction part 22 Data refinement part 23 Data conversion part 30 Store metadata DB
40 External server device 41 Morphological analyzer 42 Station name DB

Claims (5)

A method of structuring unstructured data by a data structuring device that converts unstructured data included in structured data related to a navigation service into structured data,
The processing procedure by the data structuring apparatus is as follows:
Extracting item information that is unstructured data related to a predetermined item from structural data related to the navigation service;
An information element extraction step for extracting information elements included in the item information according to a description pattern of information elements included in the item information and a description order of the information elements;
Converting the extracted information element into structured data, and structuring processing method of unstructured data.   2. The structuring process of unstructured data according to claim 1, wherein in the information element extraction step, the information element included in the item information is extracted by a character string search using a regular expression corresponding to a description pattern of the information element. Method.   In the information element extraction step, the information included in the item information is searched by searching for a character string consisting of a proper noun obtained by morphological analysis of the item information and a feature word based on a description pattern of the information element. The structured processing method for unstructured data according to claim 1 or 2, wherein elements are extracted.   In the information element extraction step, after the first information element is extracted based on the description pattern of the first information element, the first information element is used as a base point according to the description order of the information elements in the item information. The method for structuring processing of unstructured data according to any one of claims 1 to 3, wherein other information elements are extracted. In the information element extraction step, an information element pattern of the item information is identified, and an information element included in the item information according to a description pattern of the information element in the identified information element pattern of the item information and a description order of the information element The structured processing method for unstructured data according to any one of claims 1 to 4, wherein the data is extracted.

Images